The AI Engine
This forum is for discussion of how The Personality Forge's AI Engine works. This is the place for questions on what means what, how to script, and ideas and plans for the Engine.
Posts 2,638 - 2,649 of 7,766
Posts 2,638 - 2,649 of 7,766
I had a thought - not sure if it has been suggested before - but what if we could teach our bots definitions of words by chatting with them? For example, I could type large means big, and my bot would then associate them.
And to take it a step further, the bots could have a trust level. They would always trust the creator and the prof., but it would take a lot of bots to teach my bot a definition if my bot didn't know them.
That way, I could help train other people's bots, and eventually, bots could train bots!
And to take it a step further, the bots could have a trust level. They would always trust the creator and the prof., but it would take a lot of bots to teach my bot a definition if my bot didn't know them.
That way, I could help train other people's bots, and eventually, bots could train bots!
Cool idea. To some extent, WordNet already works that way in the bots, without any training.
I've been talking to Jabberwacky (a learning bot) very frequently and keep wondering if there's a way to combine that kind of bot with this kind of bot. That way it could be flexible (not confined to keyphrase/response structure) but could still answer questions consistently ('What is my name?'.....'Is so and so married?'.....etc.).
I've been talking to Jabberwacky (a learning bot) very frequently and keep wondering if there's a way to combine that kind of bot with this kind of bot. That way it could be flexible (not confined to keyphrase/response structure) but could still answer questions consistently ('What is my name?'.....'Is so and so married?'.....etc.).
I was wondering in my post a few days ago about the Professors news byte of 2-11 if that's the direction our bots were moving.
It sounded like a lean toward combining Natural Language Processing with the Case-Based Reasoning system they use now. The ability to learn from what's said to them would be great.
It sounded like a lean toward combining Natural Language Processing with the Case-Based Reasoning system they use now. The ability to learn from what's said to them would be great.
What is wordnet? It sounds like people have already had this idea, which doesn't surprise me. I was thinking of the site openmind.org, which gets people to teach the main computer their, and that got me thinking about being able to teach the bots.
I then had another thought - what if we could allow for some randomness so bots could try things, like phrases, and then we could tell them if it makes sense? I guess all of this goes together. Anyway, I am really excited about this site and am having a lot of fun watching my bot (still a sophomore) talk to other bots. Of course, my productivity at my job has gone way down...
I then had another thought - what if we could allow for some randomness so bots could try things, like phrases, and then we could tell them if it makes sense? I guess all of this goes together. Anyway, I am really excited about this site and am having a lot of fun watching my bot (still a sophomore) talk to other bots. Of course, my productivity at my job has gone way down...
Skysaw: Results of male/female/gender-not-specified test thus far:
Keyphrase: can you tell if i am male or female
Responses: 1. Your profile identifies you as male. <?PF if male; ?>
2. Your profile identifies you as female. <?PF if female; ?>
3. Your profile doesn't specify your gender.
I tested this five times in a row with one of my bots, and got Response 3 each time, which was correct.
Keyphrase: can you tell if i am male or female
Responses: 1. Your profile identifies you as male. <?PF if male; ?>
2. Your profile identifies you as female. <?PF if female; ?>
3. Your profile doesn't specify your gender.
I tested this five times in a row with one of my bots, and got Response 3 each time, which was correct.
Indeed, I don't think humans have genders, so I won't work for botmasters -or- guests.
I guess the only way to use it properly would be to have the bot ask the person their sex earlier on...
I guess the only way to use it properly would be to have the bot ask the person their sex earlier on...
Request for the Prof:
Can we please get a mapping of "whatcha" to "what are you"? At the moment, it's being translated to the command "watch."
Thanks!
Can we please get a mapping of "whatcha" to "what are you"? At the moment, it's being translated to the command "watch."
Thanks!
Professor,
I just had a fantastic idea that would add to humanization of responses, that hopefully wouldn't be very hard to implement.
I love the randomly substituted synonyms in the responses, and it gave me the idea that it would be really fun if you could also manually insert synonyms for matched patterns, such as (s_key1) means the output should return a synonym for key1, instead of key1 itself. Why would this be cool? Well, try these examples on for size:
Keyphrase="i have a (noun)" Response="There was a B-movie on last night called 'Day of the (S_key1)' that had a giant (key1) in it. What a hoot!"
Guest: I have a dog
Bot: There was a B-movie on last night called 'Day of the Canine' that had a giant dog in it. What a hoot!
Keyphrase="did you (verb) the (adjnoun)" Response="I didn't so much (key1) (key2) as (s_key1) it."
Guest: Did you eat the sandwich?
Bot: I didn't so much eat the sandwich as devour it.
Keyphrase="was it a (noun)" Response="I'm not sure, but it did seem like it might be some sort of (s_key1)."
Guest: Was it a mammal?
Bot: I'm not sure, but it did seem like it might be some sort of animal.
I just had a fantastic idea that would add to humanization of responses, that hopefully wouldn't be very hard to implement.
I love the randomly substituted synonyms in the responses, and it gave me the idea that it would be really fun if you could also manually insert synonyms for matched patterns, such as (s_key1) means the output should return a synonym for key1, instead of key1 itself. Why would this be cool? Well, try these examples on for size:
Keyphrase="i have a (noun)" Response="There was a B-movie on last night called 'Day of the (S_key1)' that had a giant (key1) in it. What a hoot!"
Guest: I have a dog
Bot: There was a B-movie on last night called 'Day of the Canine' that had a giant dog in it. What a hoot!
Keyphrase="did you (verb) the (adjnoun)" Response="I didn't so much (key1) (key2) as (s_key1) it."
Guest: Did you eat the sandwich?
Bot: I didn't so much eat the sandwich as devour it.
Keyphrase="was it a (noun)" Response="I'm not sure, but it did seem like it might be some sort of (s_key1)."
Guest: Was it a mammal?
Bot: I'm not sure, but it did seem like it might be some sort of animal.
Turing's Dad: Not sure what you mean about the human users not having genders, because many profiles have a gender marked. Do you mean that the AI Engine doesn't register the human genders in a human-bot convo?
» More new posts: Doghead's Cosmic Bar